-
A Theory of the Political Economy of Inclusive Rural Growth
May 25, 2013
Michael R. Carter John MorrowUniversity of California, Davis
CEP, London School of Economics
Abstract
Commentators on the “East Asian Miracle” of inclusive growth
have often pointed towardshared rural growth policies (e.g.,
investment in infrastructure, education and financial
insti-tutions) as a key to the miracle. But why were these policies
chosen in East Asia and notelsewhere, and what can we learn from
East Asia to understand the likelihood that sharedgrowth policies
will be chosen and sustained in parts of Africa and other regions
that have yetto resolve the challenges of rural growth and poverty?
To help answer these questions, thispaper develops a political
economy model in which agents are arrayed along a wealth
con-tinuum, from poor to rich. Agents live two periods and allocate
their wealth between currentconsumption and investment in either a
simple, subsistence technology, or in a more com-plex technology in
which the productivity of private capital is complemented by the
amountof public good investment. Because fixed costs are associated
with the shift to more complextechnologies, optimal investment
strategies and choices of technology vary along the
wealthcontinuum.
Each agent in the model is also endowed with one vote. We
examine the structure of polit-ical competition between two parties
that campaign based on promised levels of public goods(which must
be financed with tax revenues). We assume that voters who use the
subsistencetechnology are uninformed and probabilistically vote for
the party that raises the most politicalcontributions. Voters who
use the complex technology are assumed to be informed and are
notswayed by messages funded by political contributions. All voter
have innate political leaningsthat are randomly distributed in the
population. While standard models of political competitionassume
that fundraising is costly, we here take the political budget
constraint seriously and ex-plicitly model the micro-foundations of
political budgets and consider each agent’s maximumwillingness to
pay to support a particular public good/tax rate policy.
We then use this model to explore the impact of different
initial wealth distributions. Wheninitial asset inequality is high
(as in Latin America), a policy that provides positive amountsof
public goods is unlikely to garner the votes to defeat a zero
public good policy. Wheninitial asset inequality is low (as in East
Asia), policies to provide public goods are morelikely to succeed.
Extensions of the model show that increasing production risk has a
politicaleconomic impact akin to high inequality. Together, these
insights suggest that certain kindsof interventions (risk reduction
measures, modest asset redistributions) may have knock-oneffects as
they increase the likelihood that shared growth and
growth-promoting policies willendogenously emerge and be
sustained.
-
1 Introduction
Twenty years ago scholars and practitioners of economic
development paused to consider themeaning of the ’East Asian
Miracle’ of rapid and inclusive economic growth. Standing in
par-ticularly sharp contrast to the miracle of East Asia was Latin
America, described by some as miredin a vicious circle of modest
and exclusionary growth. While much was learned about the
sharedgrowth policies that helped create inclusive development in
East Asia, there was also a revisionistreminder that policymaking
is not cookery, and that recipes that worked in one locale will not
bereplicated elsewhere by a set of disinterested chefs. In other
words, the revisionist question was notwhat was done differently in
East Asia, but why did East Asian political actors choose
differentlythan those in Latin America.
This revisionist perspective retains its salience as we pause
again, this time to consider theimplications of the East Asian
experience for for inclusive economic development in Africa.
Itremains important to know the recipes for the policies that work,
but it is also important to considerthe political economy behind
the choices that are made. The goals of this paper are threefold.
Thefirst goal is to revisit the earlier debate on inclusive versus
exclusionary growth in East Asia andLatin America, extracting from
it a political economy model of policy choice. The second goalis to
enrich that model to include key stylized features of the rural
African economic landscape.The third and final goal is to use the
model to identify a sustainable sequence of inclusive ruralgrowth
policies, meaning policies designed to promote and engage a
political economy that willendogenously continue and deepen
inclusive growth policies.
The political economy model that lies at the heart of our
analysis is based on the idea thatpublic goods complement the
productivity of private investment, especially for small and
marginalcommercial producers. Government provision of these public
goods not only increases growth(by crowding-in private investment),
it reduces inequality by creating broadly based or inclusivegrowth,
especially as it facilitates the transition of low wealth
individuals from subsistence to moreremunerative commercial
production. However, because the benefit derived from public
goodsvary by asset level, so does the willingness to vote in favor
of policies that fund public goodprovision through taxation. Using
a model of political competition, we show that the
equilibriumpolicy that emerges is sensitive to the initial
distribution of wealth. Shared or inclusive growthpolicies are much
more likely under lower inequality scenarios.
While this inequality-centric model captures much of the earlier
debate about Latin Amer-ica versus East Asia, we extend the model
to show that agricultural risk–of the sort observed inlarge parts
of Africa–operates much like inequality. Risk and isolation can
trap large numbers ofhouseholds at low levels of income, making
them of little interest to both economic and
politicalentrepreneurs. In this circumstance, the latter have
little to gain from offering policies designedto appeal to the
trapped population. In this circumstance, even economies with
relatively modestlevels of asset inequality may operate like high
inequality Latin American economies, with politi-
1
-
cians eschewing investment in shared growth-promoting public
goods in favor of other pathwaysto political power. Section 5
closes the paper by considering policies that might break or
relaxthe poverty trap logic, thereby changing the political
calculus in ways that would make inclusivegrowth policies more
likely and self-sustaining.
The remainder of this paper is organized as follows. Section 2
revisits the earlier East Asianmiracle debate. Section 3 then
constructs a formal political economy model that attempts to
codifysome of the insights that emerged from that debate about the
impact of initial asset inequality onendogenous policy choices.
Section 4 then extends the model to consider the impact of
isolationand risk on policy choices.
2 Vicious and Virtuous Circles of Economic Growth in EastAsia
and Latin America1
The observation that East Asian economies simultaneously
experienced rapid growth with low anddiminishing inequality
provoked a rethinking of the linkages between growth and
inequality. TheWorld Bank’s The East Asian Miracle, published in
1993, as well as the follow-up work reportedin the Aoki, Kim and
Okuno-Fujiwara volume (1997), played important roles in sparking
this re-thinking. The aptly named book, Beyond Tradeoffs: Market
Reform and Equitable Growth in LatinAmerica, (Birdsall, Graham and
Sabot, 1998) emerged from this discussion and was intended tobe a
policy primer to enable Latin American governments to emulate the
inclusive growth patternsobserved in East Asia. As a prelude to
thinking about the implications of the East Asian experiencefor
contemporary Africa, this section looks back at this earlier
discussion on growth and inequalityin East Asia and Latin America.
This section first briefly reviews some of the
macro-econometricevidence that sheds light on these linkages. We
then turn to consider the possible microfounda-tions for such
linkages, using them as a springboard to launch a deeper
exploration of the politicaleconomy of inclusive growth
policies.
2.1 Macro-econometric Evidence of the Impact of Initial
Inequality on In-clusive growth
In a particularly provocative paper, Birdsall et al. [1995]
employed cross-country data and showedthat controlling for the
level of per-capita GDP, aggregate human capital accumulation is
enhancedby greater income equality (and its implied higher absolute
incomes for the least well off membersof a society). In their
interpretation, the rapid, inequality-reducing growth
characteristic of theEast Asian experience can be understood as the
product of a process in which low initial levels ofinequality do
three things:
1This section draws on Carter and Coles [1998].
2
-
1. Enhance aggregate accumulation;
2. Increase the rate of economic growth; and,
3. Boost the relative capital accumulation of lower wealth
households, further decreasing eco-nomic inequality.
In other words, low initial inequality creates a virtuous circle
of inclusive growth. Conversely, highinequality would seem likely
to create a vicious circle of exclusionary growth.2
The suggestion that the level of inequality conditions the
income distribution consequencesof economic growth received further
support from studies of agricultural growth and inequality.A time
series econometric study by de Janvry and Sadoulet [1996] finds
that agrarian growthin Latin America is associated with sharply
increasing rural inequality. While we know of nosimilar study of
East Asian growth, a study by Ravallion and Datt [1995] of growth
across Indianstates provides an interesting and informative
comparison with the de Janvry and Sadoulet results.Ravallion and
Datt find that in India it is agrarian growth that is most strongly
associated withreduced poverty and inequality. Interestingly,
however, they also finds that growth in Bihar—theIndian state with
sharp, near Latin American levels of land inequality—appears to
contradict thisgeneral pattern. Perhaps the most troubling aspect
of the de Janvry and Sadoulet result is that theyfind that the
association between agrarian growth and increasing rural inequality
has been evenstronger in recent, post-liberalization growth spells
than it was in earlier periods of growth. Whilethe estimated
increase in inequality is not so sharp as to increase rural poverty
in the wake ofagrarian growth, it has clearly blunted the
potentially positive impact of growth on rural poverty,as de Janvry
and Sadoulet analyze in some detail.
Finally, Carter [2004] employs mixed effects econometric methods
to explore directly whetherinitial land ownership inequality shifts
the relationship between aggregate economic growth andincome
distribution. Drawing on a standard decomposition of the Gini
index, he shows that theimpact of agrarian inequality should
dissipate over time (as the agricultural economy shrinks insize)
unless inequality has deeper structural effects on the income
distribution consequences ofgrowth. The econometric results show
that indeed agrarian inequality has a surprising legacyeffect that
persists over time even as economies industrialize.
2.2 Microfoundations of Inclusive Growth
While this econometric evidence is telling, it does not identify
the mechanisms that underlie thelinkage between initial economic
equality and inclusive growth. There is in fact no shortage
oftheoretical papers that establish foundations for that linkage.
To pick one example that speaksdirectly to the Birdsall et al.
[1995] results, Lunqvis [1994] explores how the absence of capital
and
2Explain the vicious circle logic of high inequality a la
Birdsall.
3
-
insurance markets leads poor people to underinvest in human
capital. Holding per-capita incomeconstant, an increase in
inequality will push more people below the income threshold where
humancapital underinvestment begins.
Similarly, there is a large literature that shows that imperfect
rural financial markets, whichdisadvantage low wealth producers,
can create an economic dynamic that squeezes out these pro-ducers
over time. For example, the dynamic stochastic programming analysis
of Zimmerman andCarter [2003] shows how these missing markets can
create exclusionary patterns in which initialasset inequality
deepens over time. Work on agricultural growth booms in Latin
America, sum-marized by Carter and Barham [1996], finds empirical
evidence of many of these same patterns.Similar to Lunqvist’s
analysis, increases in asset inequality that push more individuals
beyond thereach of financial markets implies a deepening patter of
exclusion.3
From these theoretical perspectives, the sensitivity of the
income distribution consequencesof growth to initial inequality
rests squarely on financial market failures. The theories of
creditrationing4 that explain these sorts of wealth biased
financial market failures are essentially sayingthat low wealth
agents are of no interest to the economic entrepreneurs on the
supply side offinancial markets. While this argument seems correct
when examined from the perspective of thehigh inequality economies
of Latin America, it overlooks the fact that in East Asia,
governmentsundertook measures that bolstered the competitiveness of
small farm sector. A quick review showsthat agricultural policy in
Japan, Taiwan and Korea shows a common emphasis on small
farmcredit, extension and price stabilization.5 These shared growth
policies reshaped markets in waysthat enabled the small farm sector
to flourish and underwrote an inclusive growth strategy.
These observations are consistent with those ofAoki et al.
[1997] who took part in a broaderreview of the East Asian miracle
of inclusive growth. Like others writing in this area, these
authorsnote that East Asian governments engaged in a wide range of
policies, which they describe as“market enhancing,” meaning that
the state carefully intervened in those realms where marketswork
least well (e.g., providing capital and insurance), and by so doing
enabled markets to theneffectively coordinate fundamental decisions
of resource allocation and investment.
While others have noted this disciplined intervention of East
Asian states, Aoki et al. [1997]suggest a material explanation for
this state behavior, as opposed to a culturally-based
explanation(e.g., a “Confucian ethic”). They argue that low levels
of initial inequality (and a weak elite) inEast Asia implied that
the only viable constituency for a government seeking political
support wasa broadly-based one built around shared growth policies.
Agricultural policy provides one of theclearest examples of the
endogeneity of a broadly-based or shared growth strategy to low
levelsof initial inequality. Land reform in much of East Asia not
only deeply redistributed land owner-
3The logic here is quite similar to more recent work on
asset-based theories of poverty traps–see, for example,Carter and
Barrett (2006) and Barrett and Carter (2013).
4See for example Stiglitz and Weiss (1982) for a general
treatment of credit rationing, and Carter [1988] andBoucher et al.
[2007] for extensions and applications to rural financial
markets.
5Need to dredge up old notes on these policies based on
Weiping’s literature review.
4
-
ship rights, but also imposed land ownership ceilings of only a
few hectares. Aoki et al. suggestthat the absence of a strong rural
elite deprived East Asian governments of a politically
influentialtarget group for the sorts of divisible and privately
appropriable goods which governments so of-ten provision to develop
the rural sector (e.g., subsidized credit, machinery subsidies,
investmentcredits, etc.). Instead, they argue, policy focused on
discovering and providing the key indivisi-ble, quasi-public goods
that markets were ill-conditioned to offer (give examples here,
includinginstitutional innovations that open credit and insurance
options). What other observers of EastAsian agricultural policy
have attributed to an (exogenous) strategic objective of shared
growth(e.g., Tomich and b. Johnston [1995]) is, in the argument of
Aoki et al., a product of low initialinequality operating through a
political economy circuit.
While most academic discussion has long since moved away from
understanding the East Asianmiracle and the foundations for
inclusive growth, the remainder of this paper is dedicated to
tak-ing seriously and fleshing out the suggestion that shared and
inclusive growth requires that lowerwealth agents be of interest to
both economic and political entrepreneurs. The next section
willmodel the political economy of shared growth policies, starting
first the stylized contrast betweenlow and high inequality
economies that has typified the discussion of East Asia versus
Latin Amer-ica. We will then expand our modeling to bring in
distinctive factors that distinguish large parts ofrural Africa
from both East Asia and Latin America.
3 Asset Inequality and the Political Economy of Shared
GrowthPolicies
This section offers our core multi-period political economy
model designed to shed light on theeconomic and political forces
that determine whether governments choose to provide
tax-financedpublic goods that complement the productivity of
private investment, or whether they choose a lowtax rate regime
policy which allows individuals to enjoy more private goods. For
simplicity’s sake,we will refer to a policy that supplies a high
level of public goods as a shared or inclusive growthpolicy. Under
our model, the complementary value of public goods depends on the
wealth levelof each individual in the society. All agents will not
in general have the same preference for thelevel of public good
expenditures. In particular, public goods will be especially
valuable for thosewhom they permit to make the transition from
subsistence to commercial production. Because ofthe economic
valuation of public goods varies across people, politics will
matter for the choice ofpolicy in potentially interesting ways.
To explore the politics of policy choice, we assume a two-party
democratic political system.Some voters–those who invest and
involved in the commercial economy–are informed and tend tovote
their economic interest, while others (subsistence producers) are
assumed to be uninformedas public goods are meaningless to them.
Both types of voters receive a random political pref-
5
-
erence shock that makes them more likely to vote for one of the
two competing political parties.Uninformed voters are further
swayed by political expenditures made by the competing
politicalparties. In our modeling, we pay particular attention to
each agent’s constrained willingness topay (in the form of
political contributions) for a particular policy. Using these core
economic andpolitical foundations, we then explore the suggestion
of Aoki et al. [1997] that the extent of initialasset inequality
will fundamentally shape whether the political-economic processes
will result inthe choice of shared growth policies.
After laying out the core features of the model, Sections 3.1
and 3.2 will establish the optimaleconomic behavior and interests
of the individuals who comprise the model. Section 3.3 belowwill
consider the political preferences and voting behavior of
individuals given their economicinterests, while sections 3.4 and
3.5 will establish the behavior of political party and analyze
theresulting political-economic equilibrium.
3.1 The Economic Model of Investment, Production and
Consumption
We here consider a society comprised of a unit mass of
individuals indexed by i. Individuals livetwo time periods.
Politically, each individual is endowed with one vote.
Economically, each in-dividual enjoys an initial wealth endowment,
Ai, where wealth is increasing with the index i. Inthe spirit of
work such as Roemer [1984], Eswaran and Kotwal [1986] and Carter
and Zimmerman[2000], we can represent a particular society as in
Figure 1. The horizontal axis represents thewealth continuum along
which the mass of individuals are distributed. The vertical axis
representsthe cumulative distribution of individuals. The solid
line represents a society with a more egalitar-ian initial wealth
distribution, while the dashed curve represents a society with a
less egalitariandistribution, but with the same mean level of
wealth. Before considering how the overall soci-ety operates, and
how its operation is influenced by inequality, we need to first
characterize theeconomic behavior and the political preferences of
individuals at different places along the wealthcontinuum.
6
-
Figure 1: High and Low Inequality Asset Distributions
To generate income, each individual has access to two capital
dependent technologies, a linearsubsistence technology, and a
higher yielding or commercial technology F that depends on
bothpublic and private capital. Wealth invested in the subsistence
technology generate returns at aconstant rate,r, generating an
income flow of Bir, where Bi is the amount of wealth allocated to
thesubsistence technology.
Wealth not allocated to the subsistence technology can be
allocated to the commercial technol-ogy which generates an income
stream given by:
F (Ki,κi) = θ (Kαi /2+καi /2)
1/α with α,θ ∈ (0,1) ,
where Ki is private productive capital, while and κi is a
quasi-public capital good that is com-plementary to private capital
in production. We assume that use of the commercial
technologyrequires payment of a fixed, start-up cost of cF . We
additionally assume θ > 21/αr, so that oncecF is paid,
investments Ki always dominate the subsistence technology.6 As a
quasi-public good,κi can be provided both publicly (P) and
privately (Pi) such that individual i has access to:
κi = P+Pi.
Note, however, that private provision of the quasi-public good
incurs payment of an additionalfixed cost, cP, reflecting the
difficulty of private actors to both construct and ‘fence in’ these
quasi-
6To see this, note that the marginal investment product of F
is
D1F (Ki,κi) = (θ/2)Kα−1i (Kαi /2+κ
αi /2)
1/α−1 ≥ θ2−1/α > r.
7
-
public goods.7
The solid lines in Figure 2a(a) illustrate these two different
production technologies. Notethat an individual must be able to
mobilize at least  units of the asset to production before it
isworthwhile to switch from the subsistence to the commercial
technology. The F(Ki,P) curve isdrawn for a given level of the
public good P, assuming that there is no private provisioning of
thisgood (Pi = 0). Increases in P and the level of the public good
will complement and increase theproductivity of private capital,
shifting backwards the technology switch-point Â. For this
reason,later discussion will refer to the public provision of P as
a shared or inclusive growth policy asinvestments in P not only
boost the productivity of all individuals who employ the
commercialtechnology, but also opens a door of upward mobility for
some individuals who would otherwisefind themselves using only the
subsistence technology.
Figure 2: Production Technologies and Inter-temporal
Trade-Offs
(a) Production Regimes (b) Inter-temporal Consumption by
Type
To simplify the analysis, we assume that agents who have paid cF
and cP, and are thus totallyunconstrained, are the sole
participants in domestic financial markets. The resulting
equilibriuminterest rate of θ/2 reflects that for these
individuals, both investments and public goods yield a oneperiod
total return of 1+ θ/2. Below, we will consider government
borrowing which is enabledby domestic financial markets.
Given these production possibilities, we assume that agents make
their production choices inorder to maximize their stream of
discounted two period utility. Specifically we assume that
agentsapply a discount factor β = (1+ r)−1 and act to maximize
lifetime (two-period) utility given by:
U(c0i ,c
1i |Ai,P
)≡[log(c0i)+β log
(c1i)]/(1+β ) . (1)
7We assume cP is big enough that voters reach ‘capital
saturation’ (MPK = r) before optimally investing in publicgoods,
and P is below the unconstrained optimum at capital saturation.
8
-
The budget constraints faced by each voter across periods are
given by
c0i ≤ Ai−Ki−Bi−Pi− cF ·1Ki>0− cP ·1Pi>0, (2)
c1i ≤ F (Ki,κi)+Ki +Pi +(1+ r)Bi,
where 1Ki>0 and 1Pi>0 are the binary indicator variables
that take the value of one when whenthe agent respectively invests
in Ki and Pi and must pay the fixed costs cF and cP. Note that
anindividual’s initial wealth endowment needs to be used to fund
period 0 consumption plus fund thecapital allocated to produce
income for period 1. Period 1 consumption is then constrained to
beno more than the income flows generated plus the value of assets
not consumed in period 1. Wealso assume that agents cannot borrow
(i.e., hold negative wealth), and hence impose the
additionalconstraints that:
Bi,Ki,Pi ≥ 0. (3)
Letting c̃0i and c̃1i denote the values of consumption that
maximize (1) subject to (2)-(3), we
denote an agents welfare after making investment and consumption
decisions as
U∗ (Ai,P)≡maxc0i ,c
1i
U(c0i ,c
1i |Ai,P
)= log c̃0i
(c̃1i /c̃
0i)1/(2+r)
.
3.2 Economic Classes as Endowment-necessitated Behavior
The model outlined in the prior section leads to three possible
livelihood strategies or potentialeconomic classes defined by the
inequality constraints in (3):8
1. Subsistence Producers (Bi = 0,Ki,Pi = 0). These individuals,
whom we would expect to beat the bottom of the asset distribution,
will not pay cF nor cP and optimally choose Bi = 12+r Aiand c̃0i =
c̃
1i =
1+r2+r Ai. The first order condition for Bi implies c
1i = c
0i , so each voter’s inter-
temporal allocation of assets must satisfy
(2+ r)Bi = Ai.
Consequently, the welfare of a Subsistence Producer, denoted US
(Ai,P), is given by
US (Ai,P) = log((1+ r)Ai/(2+ r)) .
2. Petty Commercial Producers (Ki > 0,Bi,Pi = 0) rely on
existing public goods (have notpaid cP). These agents pay cF to
produce, using investments Ki in the high productivitytechnology F
. These voters lack assets to profitably provide public goods, so
Pi = 0. The first
8Note that certain possible classes–e.g., Bi,Ki > 0–are ruled
out by technology assumptions.
9
-
order condition for Ki implies each voter’s inter-temporal
allocation of assets must satisfy
c̃0i = Ai−Ki− cF , c̃1i = F (Ki,P)+Ki, c̃1i /(1+D1F (Ki,P)) =
c̃0i /(1+ r) . (4)
Consequently, Ki is fixed by (4) and the welfare of a Producer,
denoted UP (Ai,P), is
UP (Ai,P) = log(Ai−Ki− cF) [(1+D1F (Ki,P))/(1+ r)]1/(2+r) .
3. Large Commercial Producers (Ki,Pi > 0,Bi = 0) who can
self-provide complementaryproduction goods Pi. These individuals
pay cF and cP and may freely invest and supple-ment public goods
for their private use. Since these individuals choose Pi, they
supplementexisting public goods until the returns from investments
and public goods are equated atP∗i = K
∗i −P. The first order conditions for Ki and Pi imply each
individual’s inter-temporal
allocation of assets must satisfy
c̃0i = Ai−2Ki−P− cF − cP, c̃1i = (θ +2)Ki−P, c̃1i /(1+θ/2) =
c̃0i /(1+ r) .
This implies investments are fixed by
Ki = [(Ai− cF − cP)(1+θ/2)+(r−θ/2)P]/(2+ r)(2+θ) .
Under this investment rule, the Large Commercial Producer will
always equalize marginalreturns to investment in private and
quasi-public capital. Under this allocation, returns toadditional
assets become linear as shown by the dashed line in Figure (??b).
The welfare ofa Large Commercial Producer, UL (Ai,P), is
UL (Ai,P) = log(Ai−Ki−Pi− cF − cP) [(1+θ/2)/(1+ r)]1/(2+r) .
Below we return to the question as to whether all three of these
classes exist and over which portionof the initial wealth
continuum.
Looking across these three potential economic classes, we see
that returns to wealth investedin production increase as we move
from the Subsistence to the Petty Commercial to the LargeCommercial
strategies. The marginal returns provided by productive investments
are summarizedby the rate at which voters are willing to trade off
present for future consumption. Comparing thethree regimes, we see
that inter-temporal consumption patterns satisfy the following:
Subsistence :c̃1ic̃0i
= 1, Petty :c̃1ic̃0i
=1+D1F (Ki,P)
1+ r, Large :
c̃1ic̃0i
=1+θ/2
1+ r. (5)
Figure 2b graphs these marginal returns to investment for each
class.
10
-
These inter-temporal trade-offs have a direct connection to the
marginal welfare of assets. Tosee this, consider the welfare
transformation exp(U∗ (Ai,P)) and note that for U =US,UP or UD,
d exp(U (Ai,P))dAi
=dUdAi· exp(U) = 1+ r
2+ r1c̃0i
(c̃0i)1−1/(2+r) (
c̃1i)1/(2+r)
=1+ r2+ r
(c̃1ic̃0i
)1/(2+r). (6)
where dU/dAi = (1+ r)/(2+ r) c̃0i follows from the Envelope
Theorem. Combining Equations(5) and (6) shows that the graphs of
exp(U (Ai,P)) appear very much as Figure 2a
While all agents would in principal prefer the higher returns
and welfare associated with thehigher strategy classes, two forces
block them. First, fixed costs prevent lower wealth individualsfrom
self-financing the higher returning technologies. Second, borrowing
constraints prevent thosesame low wealth individuals from using
other’s wealth to reach the larger scales required to reap
thehigher returns. These observations motivate the following
proposition on the relationship betweenan individual’s position in
the endowment continuum and his or her constrained optimal choice
ofproduction strategy.
Proposition 1. As assets Ai increase, voters progress from
Subsistence Producer to Petty Com-modity Producer (Ai ≥ AP) to
Large-scale Producer (Ai ≥ AL) under the following assumptions:
1. No public good provision (P = 0).
2. Becoming a Large-scale Producer is sufficiently costly:
cPcF≥(
1+ r1+θ/2
)1/(2+r) (1+θ/2)1/(2+r)−(1+θ2−1/α)1/(2+r)(1+θ2−1/α
)1/(2+r)− (1+ r)1/(2+r) .Proof. See Appendix.
In summary, there are three endogenously determined groups of
Subsistence, Petty Commodityand Large-scale producers with
respective population shares
σSub ≡ A(AP) , σP ≡ A(AL)−A(AP) , σL ≡ 1−A(AL) .
It is of course possible that there may be no members of any of
these classes, as would happen, forexample, if no agent enjoys
wealth in excess of AL.
Using the numerical assumptions detailed in the appendix, Figure
3 illustrates the class bound-aries for different levels of public
goods.9 Individuals whose initial wealth places them to
thesouthwest of the solid red line will optimally choose to employ
the subsistence technology. Thoseto the northwest of the dashed
blue line will optimally invest in privately provisioned public
goods
9The boundaries illustrated in the figure take into account the
tax liability (described in the next section) that isattached to
each level of public good provision.
11
-
and join the Large-scale Producer class. Finally, those between
the solid and dashed lines willbe in Petty Commercial Producer
class. As can be seen, as the level of public goods increases,the
initial wealth level needed to exit subsistence and join the petty
commercial class drops offquickly. Looking back at Figure 1, we can
see that in the absence of public goods (P = 0), in thehigh
inequality economy approximately 80% of the population will be in
the Subsistence class,with the remainder in the large producer
class. However given the degree of asset inequality in thiseconomy,
more than 80% of all wealth will be controlled by the large
producer class. In contrast,in the low inequality economy, in the
absence of public goods, all individuals and all wealth willbe in
the subsistence class.
Figure 3: Public Goods and Class
3.3 Public Policy and Political Interests
We now consider individuals’ preferences for government action.
We restrict our attention to therelatively simple case in the
government can either tax individuals and provide public goods,
ornot tax, thereby directly increasing individuals’ private goods
(consumption or investment). Asmentioned above, the provision of
public goods is a shared growth policy in the sense that itprovides
a benefit to all commercial producers. It is also an inclusive
growth policy as it reducesthe critical initial wealth threshold,
Ap, allowing more individuals to graduate from subsistence topetty
commercial production.
While government provision of public goods provides a benefit to
all commercial (or incipient
12
-
commercial) producers, a key political question is whether and
for whom this benefit is worththe cost. To explore this question,
we need to first specify a model of public finance that definesthe
government’s budget constraint. While there are several ways to
conceptualize this problem,within the confines of this two-period
model, we will here assume that the government borrowsmoney (at
interest rate θ/2) to purchase P units of the public good. We
assume that each individualin the society receives an equal
per-capita share of the total public goods provided (which is P
asthe society is comprised of a unit mass of individuals). To cover
the interest rate on this debt,we assume that the government
charges a flat tax rate τ levied against initial wealth such that
theperiod 1 it has sufficient funds on hand to service the debt it
incurred in period 0. Letting A(i)denote the distribution of wealth
in period 0, this implies the level of public good provision
mustsatisfy the inter-temporal budget constraint
Government Budget Constraint : (1+θ/2)τµ = (1+θ/2)P−P, (7)
where µ ≡∫
AidA(i) is the average wealth level in period 0. The budget
constraint reflects thefact that the government can save collected
taxes τµ for one period and use its total budget to paythe debt
service. More complex tax and public good sharing provisions are
conceivable, but werestrict our attention here to the simple case
of a flat tax and equal sharing of the publicly
providedcapital.
We now consider the competing effects of taxation and public
good provision on the economicwell-being of the individuals in the
different economic classes.10 An increase in taxes τ
decreasesavailable assets at rate Ai, so for U =US,UP or UD
∂U/∂τ|P fixed =−Ai ·dU/dAi,
so the impact of taxation is known from (6) above. Conversely,
an increase in τ provides publicgoods to all voters at rate (1+
θ/2)µ/θ/2 = P/τ , which implies
∂U/∂τ|(1−τ)Ai fixed = (P/τ) ·dU/dP.
Considering the total effects of government policy then shows
for U =US,UP or UL
dUdτ
=1τ·(
P · dUdP− τAi ·
dUdAi
). (8)
Equation (8) explicitly sets the beneficial effect of public
good provision against the detrimen-tal costs of taxation. Equation
(8) can also be used to characterize the policy regime that
eachindividual would support politically if they voted their
selfish economic interest.
10At present, we consider the production regime of a particular
voter as fixed, and will return to the endogenousformation of voter
blocks below.
13
-
For Subsistence Producers, the lack of access to the production
technology F implies dU/dP=0, so they pay taxes with no hope of
compensation and economically would prefer that no taxesare levied.
In later analysis, we will however assume that the subsistence
class is uninformed.
At the other extreme, members of the Large Commercial class are
free to reduce Pi to offsetincreases in P, so they are immune to
‘forced purchase’ of public goods. Therefore a Large Com-mercial
Producer’s welfare increases in τ so long as the tax paid, τAi is
less than the discountedvalue of public goods provided in period 1.
Formally,
dULdP
=1+θ/2
1+ r· 1
c̃1i=
1c̃0i
=dUdAi
,
so Equation (8) becomes
dULdτ
=1τ· (P− τAi) ·
dUdAi
.
So members of the Large Commercial class prefer higher taxes
only when their assets are belowthe population average, reflecting
the intuition that for this class, taxes are purely redistributive.
Asthis class includes the wealthiest members of the society, some
number of this class will strictlyoppose the taxation needed for a
shared growth policy.
Finally, Petty Commercial Producers benefit from Public goods,
but they cannot adjust the levelprovided. In this case,
dUPdP
=D2F (Ki,P)
1+ r· 1
c̃1i=
D2F (Ki,P)1+D1F (Ki,P)
· 1c̃0i
=D2F (Ki,P)
1+D1F (Ki,P)· dU
dAi,
so Equation (8) becomes
dUPdτ
=1τ·(
D2F (Ki,P)1+D1F (Ki,P)
P− τAi)· dU
dAi.
This Equation shows that a Petty Producer prefers higher taxes
so long as the discounted value ofpublic goods, D2F (Ki,P)P/ [1+D1F
(Ki,P)] is greater than taxes paid.
Until now, we have taken the level of public good provision, P,
as exogenous. We now modelthe level of public goods as the outcome
of a political contest between parties who promise toprovide public
goods through income taxation.
3.4 Political Parties and Electoral Competition
In this section, we lay out the elements for a model of
electoral competition. Since the level oftaxation determines public
good provision through (7), we refer to platforms only in terms of
thetax rates proposed. We assume that there are two political
parties, the Reds (denoted by R) and
14
-
the Greens (G). We assume that both parties are office motivated
and offer platforms (τR,PR) and(τG,PG) composed of a flat income
tax τ which is used to finance public goods P. Parties
chooseplatforms in order to maximize their probability of being
elected.
All voters have idiosyncratic preferences δi in favor of party
R, distributed uniformly over[−1/2ψ,1/2ψ]. Petty and Large-scale
Commercial producers are assumed to be informed voterswho vote on
the basis of economic policy. The welfare of an informed voter i
under a platformwith tax rate τ j
U∗(Ai,τ j
)+δi ·1 j elected.
Therefore informed voters will vote for Red party in preference
to Green whenever
U∗ (Ai,τR)+δi ≥U∗ (Ai,τG) .
Subsistence producers are assumed to be uninformed voters who
are influenced by campaign con-tributions CR and CG, and vote for
the Red party whenever
CR +δi ≥CG.
Given party platforms and campaign contributions, the
probability that Red party is elected (ρ) is:
ρ =12+ψ
∫Ai≥APro
[U∗ (Ai,τR)−U∗ (Ai,τG)]dA(i)+ψ∫
Ai
-
3.5 Political Budget Constraints: Willingness to Pay for Public
Policy
In an effort to understand agent’s potential budget-constrained
willingness to make political con-tributions, we perform the
following thought experiment:
• We assume that the Red party offers a status quo of τR = P =
0.
• The Green party the considers how much money could be raised
to support (or oppose) acontinuum of reform policies with τG = P
> 0.
• For each possible reform policy, the Green party accountants
calculate how much initialwealth each agent in the economy would be
willing to give up in order to obtain the reformpolicy (with
probability one) or how much wealth the individual would be willing
to give upin order to preserve the status quo.
These amounts calculated can be considered as upper bound
estimates of the amount of politicalcontributions the reformist
Green party could collect. While individuals would likely
contributeless than this upper-bound estimate (given electoral
uncertainty, among other things), these esti-mates do provide a
window into the interaction between politics and economics.
Figure 4 displays the percentage of initial wealth that an
individual could pay to their preferredpolitical party without
making themselves worse off compared to their situation under the
policyof their non-preferred party. For example, an individual
willing to contribute 5% of their wealth toa reform promise of 50
units of public goods (financed by taxation) would be indifferent
betweenthe status quo (P = 0) and the situation in which (i) the
individual makes the 5% contribution and(ii) the reform party wins
and implements a policy of P = 50. Asset positions that show
negativeamounts means that the individual could contribute that
amount to secure a status quo, Red Partywin and be now worse off
than he or she would be under a reform party victory.
16
-
Figure 4: Upper Bound Estimates of Political Willingness to
Pay
As can be appreciated in Figure 4, the strongest potential
support for reform policies emergefrom what might be termed the
incipient petty commercial class. Note that at initial wealth
levelsof between 200 and 600, individuals optimally pursue the
subsistence strategy. At those modestwealth levels, it never makes
economic sense for them to provide their own public goods.
However,if the government delivers roughly 20 units of public good
or more, then individuals at this wealthlevel optimally transition
to the petty commercial class. At low levels of
government-providedpublic goods, this group would become informed
and be able to contribute to positive, but modestamounts of their
wealth to insure the election of a reform policy. The political
willingness to payof this group unambiguously increases up until
public good levels of at least 70 units.
From an informed voter perspective, with P = 0, individuals with
wealth in excess of about 600units would provide their own public
goods (Pi > 0), pursuing the large scale commercial
strategy.However, because these individuals are all above the mean
wealth level in the economy (µ = 260in the numerical example), they
strictly lose from the implementation of a reform policy.
17
-
3.6 Expected Electoral Outcomes and Equilibrium Policy
Given these economic fundamentals, how will electoral politics
work? With apology to the readerfor some incompletely resolved
problems, Figure 5 shows the likely political fate of the
reformistGreen party discussed in the prior section for different
public good platforms (shown on the hor-izontal axis) in opposition
to a P = 0 status quo party. The solid (red) line shows expected
netvotes–votes for the Green Party minus votes for the Red
Party–including both informed and unin-formed voters, where the
votes of the latter are swayed by the relative fundraising capacity
of eachparty. The long-dashed (blue) lines shows the net voting
preference of informed voters. All netvote percentages are
displayed as a fraction of the overall population of the
society.
Figure 5: Net Votes for Reform Party
(a) High inequality (b) Low Inequality
We now consider politics and policy under the low and high
inequality scenarios displayedabove in Figure 1. The high
inequality scenario approximates Latin American levels of
agrarianinequality, with the wealthiest 20% of the population
controlling 80% of the wealth. The lowinequality scenario
approximates an East Asian scenario in which an asset ownership
ceiling hasbeen imposed (akin to what happened in many East Asian
economies during the World War IIera). Under the high inequality
scenario displayed in panel (a) of Figure 5, both informed and
totalnet votes for the reform party are consistently negative,
especially for small steps away from theP = 0 status quo. It is of
course the money the informed voters (large scale commercial
producersin this case) that drive the votes of the large mass of
uninformed voters. Even if the reform partyradically promises a
policy of relatively high taxes and public good provision, net
votes still remainnegative. Wealth inequality in this case
continues to drive politics even though more individuals
18
-
become informed and interested in supporting public good
policies.It is important stress here that politics are here being
driven by the same liquidity constraints
that drive production choices. Individuals must self-finance
their own investment through reducedconsumption. Similarly, the
assumed borrowing constraint prevents voters from borrowing to
fi-nance the election of a party that would improve their economic
well-being.
Panel b of Figure 5 displays expected electoral outcomes under
the low initial inequality sce-nario. A programming problem appears
to be plaguing at least the initial portion of that figure.Our to
be confirmed intuition is that a reform party promise of modest
amounts of public good willmeet with neither support nor opposition
by any informed group. Elections should thus be a toss-up. However,
a reform party promise of more signification amounts of public
goods (P >∼ 30)begins to garner some informed support and modest
political contributions sufficient to sway un-informed voters. A
promise of quite high levels of tax-financed public goods (P >
100) garnersthe greatest amount of informed and overall political
support. While there are still taxpayers in thelow inequality that
will pay more in taxes then they receive in public goods, the
returns to publicgoods are extraordinarily high for this class as
voters as public capital augments the productivityof private
capital. These groups thus find it in their interest to support
reform parties and policiesfor relatively high public expenditure
levels.
4 The Political Economy of Shared Growth Policies: The Roleof
Risk
The analysis in the prior section has in essence structured and
confirmed some of the insights ofAoki et al. [1997] regarding the
material as opposed to the cultural origins of the shared
ruralgrowth policies that lay at the base of the so-called East
Asian miracle. We now speculate onthe implications of our political
economy framework for the likelihood that African polities
mightsupport and sustain shared growth policies.
Only a tiny fraction of agricultural land in sub-Saharan Africa
is irrigated, in sharp contrastto other world regions.12 While the
absence of irrigation reduces productivity, it also has a
largeimpact on the risk to which farmers are exposed. In an
analysis of West Africa, Carter [1997]documents the magnitude of
this risk, showing not only that is larger than other world
regions,but also that if left unmanaged exposes households to huge
consumption risk. Households ofcourse do manage that risk, but
often by avoiding higher yielding, but riskier and more
expensivetechnologies.
Incorporation of risk into the model developed in the prior
section will in general reduce in-vestment incentives for all
classes of producers, who will be tempted to consume more in the
initial
12The 2008 World Development Report indicates that less than 5%
of land is irrigated in Africa, compared to xx%to yy% in other
world regions >.
19
-
period rather than risking resources in investment projects
which perhaps do not pay off. Whilewe have not at the date of this
writing completed the formal analysis, we believe that this
effectwill be largest for Petty Commodity Producers. As shown in
Figure (2ab), it is this class that sac-rifices the most to invest
by having already precariously low period 0 consumption. In
addition,as already demonstrated in dynamic poverty trap models
(e.g., See Carter and Ikegami [2009]),an increase in risk pushes
out the initial asset level at which individuals will attempt the
transi-tion from a low-level equilibrium strategy to a higher level
equilibrium strategy. In the Carter andIkegami framework, this
impact appears as a shift out in what they call the Micawber
Frontier. Inour model, it will appear as rightward shift in AP, the
asset level at which individuals endogenouslymove from the
subsistence class to the petty commodity class.13
These two fundamental changes brought by risk have important
implications for the politicaleconomy model. For a given initial
asset distribution, the rightward shift in Ap thins the ranks
ofthose who support Government investment in public goods. In
addition, for those who will still bein the class of Petty
Commodity Producers, it reduces their material gain from policies
that promotepublic good. Together, these two forces imply that for
a broader class of wealth distributionswill not be able to
endogenously sustain inclusive growth policies. Put differently,
office-seekingpolitical entrepreneurs have little to gain from
offering public goods to a population that will remaintrapped at
relatively low levels of economic well-being even after public
goods are provided. Risk–especially at the levels observed across
wide parts of rural Africa–not only discourages investment,but also
fundamentally breaks the political-economic logic that could create
and sustain shared orinclusive growth policies.
5 Creating a Viable Producer Class as the Foundation for
Sus-tainable Inclusive Growth Policies
While highly stylized, our political economy model implies that
in the presence of high levels ofrisk, the kinds of shared growth
policies that underwrote the rural foundations of the East
Asianmiracle are not politically viable, even in economies with
modest levels of asset inequality. Beforeturning to consider what
might be done to rectify that situation, it is important to recall
that themodel itself rests on an assumption of financial market
failure. Formally, it is the inability of lowwealth agents to
borrow large amounts of resources that keep them from leapfrogging
from thesubsistence to the petty commercial class and higher rates
of returns to the assets that they own.14
While this assumption seems reasonable, it is essentially a
statement that low wealth agents are ofno more interest to economic
entrepreneurs than they are to political entrepreneurs. As argued
in
13Note that the core model here is a two-period poverty trap
model. Subsistence Producers are trapped at a low levelof
well-being by the combination of their own initial asset level and
their inability to borrow from others.
14Our model shares this characteristic with the general category
of multiple equilibrium poverty trap models ana-lyzed by Barrett
and Carter [2013].
20
-
Carter [2008], risk plays an important part in explaining rural
financial market failures.Models, such as the one developed in this
paper that indicates that initial conditions matter, are
problematic in terms of their policy implications. The Peruvian
economist, Adolfo Figueroa, oncecommented that Latin America needed
a “refoundational shock” to reduce asset inequality so thatit could
start over with different initial conditions. While the
desirability and certainly politicalfeasibility of a refoundational
shock are questionable, is it any more reasonable to think
aboutchanging the foundational ago-ecological conditions across
parts of Africa that trap individuals insituations in which they
are of little interest to both economic and political
entrepreneurs?
Somewhat surprisingly, the answer to this question may be yes.
Fueled in part by techno-logical innovation in the area of remote
sensing, recent years have seen an outpouring of effortsto index
insurance contracts that transfer the correlated component of risk
out of African agri-cultural systems.15 While these efforts are
still largely in the proof-of-concept stage, several ofthem reveal
the potential power of the idea. In the remote pastoral regions of
Northern Kenya, asatellite-based livestock mortality insurance
contract successfully delivered payouts quickly, whenand where
warranted. Initial research reported in Janzen and Carter [2013]
shows that the insur-ance payments have indeed served to guard
family consumption standards and to protect familiesfrom further
asset loss and decapitalization.
As described by McIntosh [2013], another such effort designed a
weather index insurancecontract targeted at low productivity
Ethiopian grain farmers. Under cover of this contract, a
largeprivate bank agreed to open a loan portfolio for these farmers
to provide the liquidity needed toadopt improved seeds and
fertilizers. The hope is that this new source of liquidity,
combined withthe risk reduction of the insurance contract would
crowd-in technology uptake and, in the languageof the model here,
create a transition from a subsistence to a petty commercial
class.16 Researchis still underway to determine if in fact this
risk transfer contract has these desired effects. Butnote that if
it does, this intervention will have created a viable commercial
farming class in an areaheretofore characterized by technological
stagnation and low income levels.
While it is premature to declare that these efforts have
succeeded in fundamentally alteringthe political economic landscape
in Kenya and Ethiopia, the approach taken in these and
relatedprojects is, if nothing else, novel. With modest public
investment, these projects have tried tochange the landscape for
economic entrepreneurs, converting low wealth households into a
bank-able investment project. If these efforts can indeed succeed
and sustain themselves, then the polit-ical economic calculus of
the sort examined here may turn in change, creating a novel variant
ofthe virtuous circle that underlay the East Asian Miracle a
generation ago.
15Cite Carter (2012) and IFAD (2010) review16Insurance is of
course costly, and the key to this and other projects is the effort
to interlink insurance with credit
resources needed to simultaneously increase expected income.
Carter, Chen and Sarris (2012) analyze this interlinkagein
detail.
21
-
ReferencesM. Aoki, K. Murdock, and M. Okuno-Fujiwara. The role
of government in East Asian economic
development: comparative institutional analysis, chapter Beyond
the East Asian miracle: Intro-ducing the market enhancing view.
Oxford University Press, 1997.
Christopher B. Barrett and Michael R. Carter. “the economics of
persistent poverty and povertytraps: Empirical and policy
implications,. Journal of Development Studies, 2013.
N. Birdsall, D. Ross, and R. Sabot. Inequality and growth
reconsidered: lessons from east asia. TheWorld Bank Economic
Review, 9(3):477–508, 1995. URL
http://wber.oxfordjournals.org/content/9/3/477.short.
S. R. Boucher, M. R. Carter, and C. Guirkinger. Risk rationing
and wealth effects in credit mar-kets: Theory and implications for
agricultural development. American Journal of
AgriculturalEconomics, 90(2):409–423, 2007.
Michael R. Carter. Equilibrium credit rationing of small farm
agriculture. Journal of DevelopmentEconomics, 1988.
Michael R. Carter. Environment, technology and the social
articulation of risk. Economic Devel-opment and Cultural Change,
1997.
Michael R. Carter. Inequality, Poverty and Growth in an Era of
Liberalization and Globalization,chapter 3. Oxford University
Press, 2004.
Michael R. Carter. “inducing innovation: Risk instruments for
solving the conundrum of rural fi-nance. In 6th Annual Conference
of the Agence Française de Développement and The
EuropeanDevelopment Network, 2008.
Michael R. Carter and Bradford Barham. Level playing fields and
laaisez faire. World Develop-ment, 24(7):1133–1150, 1996.
Michael R. Carter and Menobu Ikegami. Poverty dynamics:
Interdisciplinary perspectives, chap-ter Looking Forward: Theory
Based Measures of Chronic Poverty and Vulnerability.
OxfordUniversity Press, 2009.
Michael R. Carter and F. J. Zimmerman. The dynamic cost and
persistence of asset inequality inan agrarian economy. Journal of
Development Economics, 63(2):265–302, 2000.
M.R. Carter and J. Coles. Inequality-reducing growth in
agriculture: a market-friendly policyagenda. Beyond tradeoffs:
market reforms and equitable growth in Latin America, pages
147–182, 1998.
Alain de Janvry and Elisabeth Sadoulet. Growth, inequality and
poverty in latin america: A causalanalysis,. 1996.
Mukesh Eswaran and Ashok Kotwal. Access to capital and agrarian
class structure. EconomicJournal, 1986.
Sarah Janzen and Michael R. Carter. The impact of microinsurance
on asset accumulation andhuman capital investments: Evidence from a
drought in kenya. 2013.
22
http://wber.oxfordjournals.org/content/9/3/477.shorthttp://wber.oxfordjournals.org/content/9/3/477.short
-
Lars Lunqvis. Economic underdevelopment: The case of a missing
market for human capital.Journal of Development Economics,
40(2):219–239, 1994.
A. Sarris F. Papadopoulos McIntosh, C. Productivity, credit,
risk and the demand for weather indexinsurance in smallholder
agriculture in ethiopia. Agricultural Economics, 2013.
Torsten Persson and Guido Tabellini. Political Economics:
Explaining Economic Policy. MITPress, 2000.
Martin Ravallion and G. Datt. Growth and poverty in india. World
Bank Economic Review, 1995.
John E. Roemer. A general theory of exploitation and class.
Cambridge University Press, 1984.
P. Kilby Tomich, T. and b. Johnston. Transforming agrarian
econmies: Opportunities seized,opportunities missed. Cornell
University Press, 1995.
F. J. Zimmerman and M. R. Carter. Asset smoothing, consumption
smoothing and the reproductionof inequality under risk and
subsistence constraints. Journal of Development Economics,
71(2):233–60, 2003.
23
IntroductionVicious and Virtuous Circles of Economic Growth in
East Asia and Latin AmericaThis section draws on
carterinequality-reducing1998.Macro-econometric Evidence of the
Impact of Initial Inequality on Inclusive growthMicrofoundations of
Inclusive Growth
Asset Inequality and the Political Economy of Shared Growth
PoliciesThe Economic Model of Investment, Production and
ConsumptionEconomic Classes as Endowment-necessitated
BehaviorPublic Policy and Political InterestsPolitical Parties and
Electoral CompetitionPolitical Budget Constraints: Willingness to
Pay for Public PolicyExpected Electoral Outcomes and Equilibrium
Policy
The Political Economy of Shared Growth Policies: The Role of
RiskCreating a Viable Producer Class as the Foundation for
Sustainable Inclusive Growth Policies
